Users often decide what media assets to consume and avoid based on the content type of the media asset. While conventional systems typically classify the content of media assets as a whole (e.g., plot summaries of programs, genre tags, etc.), this coarse-grained approach may oversimplify the varied content types within the media asset. Under these systems, users wishing to avoid undesirable segments and target desirable segments lack effective mechanisms for distinguishing between different types of content across different segments of a media asset.